Prediction the Ground Water Level and Deformation of Adjacent Foundation for Saturated Sand Soil by ANN Approach

Geotechnical engineering is important, as can be seen in excavation sites for new buildings being built next to existing ones. It is a major problem that population density in urban cities was on the rise, leading to an actual crisis in housing and vertical building constructions at city centers demanding stability of foundations and soil hydraulics studied and investigated with various soil types. This research studied the influence of hydraulic properties of saturated sandy soil using experimental method for simulating the fields, a laboratory device with three gradations sandy soil saturated, a (135) experiments were carried out with different parameters, varying excavation depth next to sheet pile, distance between foundation and sheet piles, and apply three vertical loads on the shallow foundation besides measuring both foundation settlement and water table downstream. including practical experiments for each type of soil. Results confirmed that as the intensity of excavation was increased, both the foundation displacement and water level rose; these results gradually decreased with finer soils; meanwhile, as far as a distance between adjacent foundation and sheet pile is concerned, it has a reverse proportionality to displacement and water level—so smaller the equilibrium status appears critical—so larger the distance from adjacent structure. As for the context applied to stress on shallow foundation, the correlation is directly proportional with the increment of value of stress, foundation displacement respectively, and incremental water level. The prediction of soil behavior and stability of foundations was performed using the statistical analysis for experimental data, which is a proposed method using artificial intelligence technique based on weight in representation by neural networks. approaches using artificial intelligence models can accommodate the mathematical relationships between independent and dependent phenomenon variables with sufficient fidelity to yield predictive solution structures accurately. The results of the analysis showed that reasonable prediction accuracy was assumed by the artificial intelligence model; for the same outcome, 97% and 98% were regression coefficients related to the artificial intelligence model, respectively.